Handle errors on system upgrade

Using 'wait' will not return error if any of the
child background process fails. This results in
unexpected behavior of triggering upgrade run
eventhough system upgrade failed. Observed this
issue during the integration of NFV's NIC partitioning
feature, in which system upgrade hangs on SR-IOV
cleanup step.

Used `wait $PID` to handle the states of all child
processes and wait for all to complete. Exit with error
if anyone of the childs exited with error.

Change-Id: I084f3b974e37c9f5a4d720747e408645deb50995

templates/fast-forward-upgrade/overcloud_system_upgrade.sh.j2

@@ -34,6 +34,7 @@ echo "[$(date)] Started system upgrade step for {{ hosts }}"
 
 tripleo-ansible-inventory --stack {{ overcloud_stack_name }} --static-yaml-inventory upgrade_inventory.yaml --undercloud-connection ssh --undercloud-key-file /var/lib/mistral/.ssh/tripleo-admin-rsa --ansible_ssh_user tripleo-admin
 
+bkg_pids=""
 for host in $(echo "{{ hosts }}" | sed "s/,/ /g")
 do
   openstack overcloud upgrade run ${RUN_ANSWER} \
@@ -41,8 +42,19 @@ do
           --static-inventory upgrade_inventory.yaml \
           --tags system_upgrade \
           --limit "${host}" 2>&1 | tee -a "RHEL_upgrade_${host}" &
+  bkg_pids+=" $! "
 done
 
-wait
+status=0
+for p in $bkg_pids; do
+    if ! wait $p; then
+        status=1
+    fi
+done
 
-echo "[$(date)] Finished system upgrade step for {{ hosts }}"
+if [[ $status == 0 ]]; then
+    echo "[$(date)] Finished system upgrade step for {{ hosts }}"
+else
+    echo "[$(date)] Failed in system upgrade step for {{ hosts }}"
+    exit 1
+fi